“Lights on, Lights off”

Evaluating outages, voltage, and economic impacts of grid investments in Accra, Ghana

The Center for Effective Global Action
4 min readAug 12, 2022


In this post, researchers Susanna Berkouwer (Wharton School, University of Pennsylvania), Geetika Pandya (Haas School, UC Berkeley), and Steven Puller (Texas A&M) discuss results from their baseline evaluation of Millennium Challenge Corporation (MCC) and Government of Ghana grid investments under the Ghana Power Compact.

Photo taken by the GridWatch research team

The problem

Imagine a problem so pervasive in daily life that it warrants the coining of a new word to describe it. Dumsor means “lights on, lights off” in the Akan language and refers to the persistent power outages across Ghana between 2012–2015. The frequency and duration of power outages was a major business constraint and hurt Ghana’s economy. In 2013, Ghanaian businesses viewed electricity reliability as a major constraint, and reported an average of 700 hours of outages annually.

The intervention

To combat this issue and strengthen the Ghanaian electricity grid infrastructure, the Government of Ghana and the Millennium Challenge Corporation (MCC) signed the Ghana Power Compact in 2014. Its goal was to increase electricity access and reliability, while attracting private-sector investment. One of the projects under the Compact aimed to improve the quality and reliability of electricity by constructing additional transformers throughout Accra.

The evaluation

Our team, composed of economists at UC Berkeley, Texas A&M, and the University of Pennsylvania, as well as engineers from UC Berkeley and nLine, Inc., was contracted by MCC to conduct a socio-economic impact evaluation of the construction of additional transformers in Accra. To estimate the extent to which the grid investments would improve access and reliability, we designed a quasi-experimental evaluation strategy. We identified households and businesses who were most likely to be affected by the intervention (‘treatment sites’), and selected statistically-similar households and businesses as a comparison group (‘control sites’). The figure below displays the location of treatment and control sites across Accra.

75 control sites and 76 treatment sites are located across the western half of Accra. At each site, we deployed four PowerWatch devices and surveyed between ten to fifteen household and business respondents.

Our engineers developed a low-cost technology suite called GridWatch, centered around sensors that collect temporal and spatial data of a power outage and measure voltage quality. The team also built a back-end cloud computing infrastructure to collect the data in near real-time and identify space-time clusters that pinpoint power outages with high confidence. The team installed these sensors at 300 households and 300 businesses, evenly distributed across the control and treatment sites..

Baseline results

The GridWatch sensors started collecting data in 2018. Using these data, we analyzed whether they exhibited parallel trends before the intervention. The figure below displays one of our key outcome variables: average voltage experienced by households and businesses at the study areas.

Data from measurements taken by the GridWatch technology in Accra at 75 control sites and 76 treatment sites. “Bad voltage” is defined as experiencing voltage outside nominal voltage (230 V in Ghana) plus or minus 10%.

Nominal voltage in Ghana is 230V and most appliances require the voltage to be close to this target level. In countries in North America and Europe, voltage is almost always stable around this target. But as shown in the figure, Ghanaian households and businesses frequently suffer from low voltage, which we define as being outside of nominal voltage plus or minus 10%.

We also conducted a baseline socio-economic survey in collaboration with our in-country partner, the Institute Of Statistical, Social And Economic Research (ISSER) at the University of Ghana. We surveyed businesses and households in metropolitan Accra and found that most measured socioeconomic and demographic characteristics, including income, education, energy usage, and business productivity , appear similar across respondents, validating our selection strategy.

Next steps

The baseline results have given us a lot to think about. Did voltage improve in targeted sites due to the intervention? And if so, did this affect household well-being and productivity of businesses? In July 2022, we will conduct an endline survey collecting data from the same household and business respondents after an exposure period of fifteen months. The construction of transformers in targeted sites concluded in March 2021, which should have given our respondents enough time to realize the effects on their electricity usage. Our analyses in the subsequent months will compare engineering and socioeconomic measurements between sites in the post-construction period, helping MCC answer core questions regarding the efficacy of their investments. In the longer term, we hope that the information collected in this study will be used to improve the delivery of electricity in Accra.



The Center for Effective Global Action
Editor for

CEGA is a hub for research on global development, innovating for positive social change.